fix: ChatContext is not a sequence — use .items for len/list; add test suite

Bug fix:
- GemmaLLM.chat(): len(chat_ctx) → len(chat_ctx.items)
- GemmaLLM._compact_context(): list(chat_ctx) → list(chat_ctx.items)
  ChatContext in livekit-agents 1.7 is not iterable or sized directly;
  it exposes an .items list. The TypeError was silently killing all LLM
  responses (agent heard user but never spoke back).

Test suite (tests/):
- test_web_ui.py: static assets, token endpoint, LiveKit WS proxy
- test_llm_api.py: completion, tool calling, thinking-disabled latency, streaming
- test_mcp_tools.py: weather (Fahrenheit), time, memory, skills — all via
  docker exec + in-container MCP client
- test_agent_integration.py: process alive, worker registered, supervisord,
  container health, LiveKit API
- test_compaction.py: size-triggered compaction, system prompt preservation,
  short-context no-op
- conftest.py: shared fixtures (HTTPS client, token, LLM, docker exec helpers)
- Run with: .venv-tests/bin/python -m pytest tests/ -v

tests/README.md

@@ -0,0 +1,38 @@
+# Hope test suite
+
+End-to-end tests that drive the **actual running container** — no mocks.
+They run on the host and hit the live web UI, token endpoint, LLM API, MCP
+tool servers, and agent process.
+
+## Prerequisites
+
+- The container is up: `docker compose up -d` (and healthy)
+- Host deps installed: `pip install -r tests/requirements.txt`
+  (or `uv pip install -r tests/requirements.txt`)
+- The LLM at `GEMMA_BASE_URL` (default `http://192.168.86.2:8023/v1`) is reachable
+- Internet access for the weather tool tests (wttr.in)
+
+## Run
+
+```bash
+cd /home/shane/dev/voice
+python -m pytest tests/ -v            # everything
+python -m pytest tests/ -v -m "not slow"   # skip LLM/network-heavy tests
+```
+
+## What each file covers
+
+| File | Target | Notes |
+|------|--------|-------|
+| `test_web_ui.py` | nginx HTTPS UI (:8090) + token endpoint (:8091) | static assets, `/token` (direct + via nginx), `/livekit/` proxy |
+| `test_llm_api.py` | Gemma LLM API directly | completion, tool calling, `enable_thinking=false` latency, streaming |
+| `test_mcp_tools.py` | MCP servers via `docker exec` | weather (Fahrenheit check), time, memory save/recall/list, skills save/recall/list — probe files are cleaned up |
+| `test_agent_integration.py` | running container | agent process alive, worker registered with LiveKit, all 4 supervised processes up, container health, LiveKit HTTP API |
+| `test_compaction.py` | `GemmaLLM._compact_context` in-container | compaction triggers >24 items, system prompt preserved, short context untouched |
+
+## Notes
+
+- LLM and MCP tests are marked `@pytest.mark.slow` (1–30s each).
+- The token endpoint is **POST** `/token` → `{"token": "<JWT>"}`.
+- Process checks use `/proc/*/cmdline` because the container image has no
+  `ps`, and the supervisorctl unix socket is not exposed in this build.